Interpretation and Mapping for the Prediction of Sites at Risk of Landslide Disasters: From Aerial Photography to Detection by DTMs

Abstract

By grasping the slope topography three-dimensionally and typifying and mapping the landform features, it is possible to visualize the areas at risk of landslide-induced disasters. This effort began with the interpretation of aerial photographs, and, in recent years, has shown great development with the advancement of remote sensing technology. In this chapter, we introduce the characteristics that can be extracted from remote sensing data that can be used to identify potential areas exposed to mass movement processes and landslide-induced disasters. In addition, we created a manual for landslide recognition by aerial photographs, AW3D-based wide-area maps, and specific landslide microtopography maps, which can be freely used to identify landslide landforms.

1 Introduction (Basic Focus, Pointing Out Issues, etc.)

Assessing and mapping the potential of slope disasters is essential for disaster prevention. In geomorphology, we have promoted the understanding of the geomorphological characteristics of slopes and have made efforts to mitigate slope-induced disasters with this knowledge. Part of this effort was greatly expanded by the acquisition of technology to grasp the terrain in three dimensions. For example, the topographic classification of slopes by substantive visual interpretation of aerial photographs, which began half a century ago, can be shown. On the other hand, remote sensing technology has advanced rapidly in recent years. This progress is not just cutting-edge. Both sensing tools and analysis software are revolutionizing the ease of using them. In the past, cutting-edge technology was the domain of some experts. It was common for highly trained technicians to obtain results with large budgets. However, in recent years, high-precision sensing equipment can be purchased at an extremely affordable price, and free analysis software can be used to achieve this goal by general engineers and anyone interested quickly. Aerial photo interpretation techniques obtained by accumulating long-term study of thumb are sometimes evaluated as “a masterly performance.” The recent “generalization of advanced technology” related to sensing may be able to “obtain results for the time being” without a sufficient understanding of the natural phenomenon itself. As a result, there are some examples where it is difficult to say that disaster prevention maps represent reality. Perhaps there is some difference between mapping by visual interpretation of aerial photographs and mapping by utilizing digital sensing technology alone.

This report summarizes the current geomorphological understanding that contributes to slope disaster prevention. In addition, the manual covers from the interpretation of aerial photographs to creating slope disaster prevention materials using UAVs. We will not only visualize the topographical understanding by visual interpretation of aerial photographs, but also try to achieve a common understanding among the parties concerned. A three-dimensional (Hereafter abbreviated as 3D) understanding of topography will pave the way for the extraction of landslide topography by detection from the contour lines. If these experiential abilities are fostered, it will be an opportunity to realize various effects in the 3D display of DEM-based terrain.

2 Mass-Movement on Slopes and Typology of Slope Disasters

2.1 Geomorphological Understanding of Slope Mass-Movement

Typifying and understanding the slope mass-movement phenomenon is a basic procedure for natural science and landslide disaster prevention. There are two directions to this. One is for the mechanistic understanding of phenomena, and the other is for understanding their spatial distribution. Mapping is, of course, the latter. Contributing to disaster science is similar to predicting potentially unstable areas. Here, there are some points to remember when implementing slope variation and mapping from a geomorphological point of view. Starting with grasping the actual distribution of phenomena, grasping the actual distribution of slope fluctuations based on a morphological understanding of slope variation characteristics, developing it into a place-based assessment of disaster potential, summarizing them and visualizing them as a map. In doing so, it should be noted that (1) the categorization is sufficiently discussed and valid. (2) It will be possible to grasp and confirm the typified phenomenon in various situations where the actual slope area and topography are grasped. In order to grasp and map topography geomorphologically, it is necessary to understand from the following four perspectives. It is the viewpoint of shape, material composition, formation action, and formation time (Miyagi et al. 2004). Here, the shape and material composition can be grasped by observation. On the other hand, the formation process and formation time are understood as a result of consideration. Shape is an introduction to topographical understanding, and if it is considered and mapped, it will be an achievement (Figs. 1 and 2).

Fig. 1

What elements does the terrain consist of? Shadow maps using 5 m DEM can grasp the distribution of the terrain that constitutes the ground surface with an accuracy of about 1/10,000. Of course, the accuracy of the data itself must be good

Fig. 2

Let’s consider what characteristics the slope consists of terrain

The erosion effect is conspicuous by the effect of water currents that create water systems and valleys. However, slope-altering failures and landslides are also linked to valley development (Miyagi 2014).

2.2 The First Step in Understanding Slope Topography, Such As Landslides

Landslides, typical of slope fluctuations, are greatly feared by local residents because of the tremendous movement and enormous damage caused by the fluctuations. Therefore, much effort was put into elucidating the mechanism of its occurrence and formulating countermeasures. UNDRR (2015) termed the word Disaster: “A serious disruption of the function of community or society at any scale due to hazardous events interaction with the condition of exposure, vulnerability and capacity, leading to one or more of the following: human, material, economic and environmental losses and impacts”. If we apply disruption here in the context of this report, it will correspond to the slope change caused at hand. From a geomorphological point of view, it can be positioned as the latest scene of various topographic change phenomena on slopes. If we can accurately grasp the changes in the shape of the slope in this single frame, we can find “traces of similar changes” that have been carried out in the surrounding slope areas using this as a document. In this way, if we can separate and map areas where similar effects are dominant and where they are not, we can provide basic data on slope disaster risk. At least in humid areas, the current topographic change phenomenon seems to have continued during the Holocene (Ex. Nakayama and Miyagi 1984; Miyagi et al. 1995).

As the scope of human activity expands and becomes more complex, the areas with the potential for new disasters to occur will expand, and the content of the vulnerability of disasters will also change. It can be said that the present age is an era in which the technology for observing topography (land) is almost complete. If so, based on understanding geomorphological characteristics, we should systematically predict where and how the slopes caused by landslides exist, when they move, and how many disasters their fluctuations are likely to occur, and realize rational and safe land use.

In Japan, as an institutional project of the National Research Institute for Earth Science and Disaster Resilience, from 1982 to 2014, we completed the deciphering work of landslide topography throughout Japan. We made it into a GIS database and made it public. Since “landslides” are also a phenomenon of topographic deformation on the ground surface, the authors had the prospect that by carefully observing the terrain, it would be possible to grasp the actual situation of “where” and “how”. A typical example of this can be found in a sketch of the shape and structure of a landslide. For example, from Vernes (1978) to Highland and Bobrowsky (2008) through Oyagi (2007), many cases have illustrated the relationship between the shape of landslides and their mode of motion and material composition. The following three points are common to many of these figs. (1) Landslides and land have clear topographic boundaries between the surrounding non-landslide lands. (2) In the land caused by landslides, there are three topographic areas: sliding cliffs, moving bodies, and, in some cases, slippery surfaces. (3) Even if a landslide causes the terrain, the micro-topographical characteristics of (1) and (2) are obscured by vegetation recovery, obedience in geomorphology, erosion due to gully extension, etc.

2.3 Synthesize and Categorize the Slope Mass-Movement Phenomena

Mass movements, which are erosion effects on slopes, are extremely diverse in shape, scale, speed, continuity of fluctuations, predispositions and incentives. Here, mass movements are summarized into three major categories from a geomorphological perspective: surface failure (shallow landslide), debris flow, and landslide. Surface failure (shallow landslide) is a mass movement that occurs under the control of the characteristics of the profile (slope longitudinal profile) from the existing ridge to the valley floor, and lumps together what is called debris flow in some regions.

The slope between the ridge and the valley floor is called the valley wall slope. Due to the shape, material composition, and supply of precipitation on the slope of this valley wall, the water table rises in a part of the slope, and an increase in pore water pressure and saturated surface currents occur, causing part of the soil mass to move. The movement of this type of slope material is called surface failure or surface landslide. Unstable materials will accumulate on the slopes of the valley walls due to various weathering and creep, and these materials will be temporarily deposited at the slope legs and valley floors. This sediment is carried and removed by occasional runoff and running water, so the slope legs are always unstable as a mass movement.

Debris flow is a phenomenon in which moving substances caused by mass movements generated in a part of a slope and sediment stored in mountain streams flow down with a large amount of outflow. Debris flows are transported and deposited differently from collapses and landslides. Surface failure and debris flow are destructive slope mass transfers on valley wall slopes. These mass transfers create distinct microtopography through destruction. However, unlike the following landslides, it does not leave moving objects on the mass transfer path of the slope.

The third type of landslide is a mass movement that is also influenced by the geological structure and occurs to deform the slope shape itself. Landslides are notable for their microtopographic composition and scale. At the same time, it is characterized by the accumulation of fluctuating unstable soil masses called mobile bodies on the slopes, forming landslide landforms. This topographic type is also diverse in shape, scale, and movement mode. Still, the unity is clearer than the previous two. This mass movement is a topographic unit separated from the surrounding slope by a clear inclination turning line. This mosaic consists of three topographic units: the main scarp, the moving body, and the sliding surface. The main scarp forms the boundary between the landslide area and the surrounding slopes. The slippery surface is exposed depending on the exercise, such as a slump. The mobile body is a fluctuation area caused by a landslide, and part remains on the slope. A slippery surface bounds the immovable area. As shown in Figs. 4 and 5, the moving body results from landslide fluctuations and is a micro landform composed of preliminary materials for the next fluctuation. From the viewpoint of slope disaster prevention, Japan adopts three types of slope fluctuations: steep slope landslides, debris flows, and landslides (Fig. 3).

Fig. 3

Classification of slope disasters in Japan (Ministry of LITT, Japan 2020)

2.4 Autonomous Destruction Process by Landslides

In landslides, elementary fracture occurs on the slope, causing deformation and alteration of the internal structure of the moving body, making it vulnerable. The moving body in the landslide terrain will undergo an autonomous fracture process (Fig. 4), giving rise to new microtopographic features. Here, we focus on (2) the basic topographic characteristics of landslides mentioned above. The occurrence of primary landslides on a part of a slope is synonymous with creating new landforms on the slopes.

Fig. 4

The autonomous destruction process: Initial landslide causes secondary deformation and alteration of moving bodies and surrounding topography (Miyagi et al. 2004)

Depending on the scale of the landslide and the movement style, it is common for a part or all of the moving body to not move completely on the slip surface and stop on the slope. If a part of the slope fluctuates as a large mass, the fluctuated mobile body will be more or less crushed. This fracturing alters the mobile body’s mechanical, hydrological, and chemical properties, and some of these changes are likely to be reflected in topographic deformation. At this time, depending on the scale of the landslide deformation, the shape of the slope that occurs, the material to be deformed, the amount of moisture, etc., most of the moving body is transported, deposited on the slope leg, or washed away. In the case of landslides, the autonomous destruction process of moving soil masses generated by landslides (Miyagi et al. 2004; 2021) can be interpreted asmicrotopography (Fig. 5). Needless to say, the various deformations due to landslides described here are phenomena that should be observed three-dimensionally.

Fig. 5

Landslide deformation is caused by surface weathering, soil erosion, and development of water systems. As the landslide deformation process ends, the landslide microtopography gradually becomes obscured. (Modified from Miyagi et al. 2004)

3 Actual Interpretation of Landslide Topography and Utilization of Digital 3D Information

3.1 Development of Real Visual Interpretation of Landslide Topography

Since the landform is three-dimensional, observing it as a three-dimensional object is desirable. If you can observe the terrain from a bird’s eye view and in three dimensions, you can grasp the whole picture and details, and above all, it is directly linked to mapping. From this point of view, the actual vision of aerial photography was an extremely important goal. It is also understandable that visible aerial photographs remain military secrets in many countries. The use of aerial photography in Japan became common after the 1960s. Technological progress in grasping and expressing topography in three dimensions has been remarkable in recent years. In recent years, advances in digital 3D technology seem to have changed drastically. This change began with the emergence of highly accurate digital 3D topographic maps by acquiring LiDAR data. Globally, SRTM data from the 90-m grid by NASA in the United States has long been made available free of charge. AW3D data is being prepared by Japan’s JAXA/ RESTEC/ NTT data, and DEM data for the 30 m grid can also be used free of charge here. In addition, the Geographic Agency of Japan has made GIS information public, the development of national land numerical information has progressed, and the free use of aerial photographs and topographic maps managed by the Geospatial Information Authority of Japan has been achieved, and 5 m DEM can be used in most of the national land. Approaches such as SAR, which use various types of satellite sensor data and utilize data with various accuracy as needed, are also common. In addition to LiDAR data, the accuracy of land information by applying multi-view photogrammetry technology is also progressing. One of the features of recent technological progress is the simplification of processing technology and the reduction of the price of measuring instruments in data acquisition and subsequent visualization processing.

3.2 Considering the Significance of the Shift from Physical Vision to 3D Data Processing

We are a group of researchers and engineers related to slope disasters. We are considering how to visualize slope disaster potential by capturing topography three-dimensionally and using our respective technologies. Furthermore, we take seriously the need to realize the visualized image as a common understanding among as many people as possible. The acquisition and retention of technology will not be completed simply by creating a manual. It is naturally necessary to visualize it in three dimensions.

Reading the actual vision of aerial photographs is a great deal of difficulty. Needless to say, the image by real vision is obtained by synthesizing the data obtained by both eyes with the brain. This image can only be visualized by the reader himself. Therefore, even if a distribution map of the landslide terrain was created by transcribing the reading results on a topographic map, “such an image could be created from these aerial photographs, and the topographic characteristics of this area would be such a thing. Therefore, we have certified this area as a landslide landform.” Suppose the distribution map of the completed landslide terrain is examined in situ. The diagram accurately grasps the distribution entity. In that case, it will never lead to the acquisition of deciphering skills, even if it suggests that the reading technician is excellent.

On the other hand, the high accuracy, lower cost, and simplicity of digital 3D information brought about by various sensing tools can facilitate the 3D imaging (visualization) realized in the brain in aerial photo interpretation. Progress in this area has been extremely rapid. In 2023, like the iPhone 14 Pro, Su-Mart phones will be equipped with this function, allowing them to create three-dimensional stereoscopic images of the terrain in front of them in a crowded field of vision, and share and observe them (Fig. 6).

Fig. 6

Using the functions of iPhone 14 Pro, digital 3D information can be acquired and processed to explain the landslide deformation of familiar places in an easy-to-understand manner. It can be easily applied to various applications such as disaster prevention patrols with local residents, basic materials for regional disaster prevention charts, and on-site medical records for disaster management. (Personal data taken by Koike 2023)

4 Mapping Slope Changes Using AW3D

4.1 Mapping AW3D Data to Slope Variation

The 3D informatization described in Chap. 3 summarizes the progress of that tool. Figure 9 shows an example of slope disaster prevention mapping in northern Vietnam’s mountainous province. Topographic analyses were performed using 5 m and 10 m contour maps generated from AW3D’s 2.5 m grid DSM and DEM. The step was to (1) extract ridge and valley lines by reading contour lines’ flexion patterns to grasp spatial distribution characteristics such as slope scale and slope. This is because, as already mentioned, collapse-type topographic changes occur on the slopes of the valley walls, so the longitudinal section from the ridge to the valley floor is characterized as a basic unit. Since debris flows have a collapse source at the top and often flow down the water system, it is meaningful to obtain an aqueous system. Next, the landslide terrain area was extracted. (2) Since the landslide topography consists of a steep slope that forms a horseshoe-shaped plan called the main sliding cliff and a gentle slope that extends in front of the slope called a landslide-moving body, the location of the bending characteristics of such contour lines was extracted. (3) Using contour lines, the slope of each grid was calculated and stratified. The steep slope of Japan is designated as a dangerous slope at about 30 degrees. (4) Using Google Earth images, slopes where vegetation was destroyed by surface failure and debris flow were extracted as places where slope disasters occurred in the near past.

There were two major difficulties with aerial photography. One is that three-dimensional topography recognition through physical vision is realized in the brain, and sharing the details with others is difficult. Therefore, to educate and acquire this technology, it is necessary to realize a common understanding by verbalizing images by conducting topographic surveys and training in visual interpretation. The second difficulty is that some countries still treat aerial photographs as military secrets, and civilians, engineers, and researchers are not allowed to handle them freely.

On the other hand, digital 3D information obtained using various sensing technologies can be said to have realized the visualization of the work performed in the brain called physical vision. It is also necessary to note that the characteristics of the sensing tool itself vary, and its mechanical accuracy and difficulty in acquiring data depend on the land cover situation and cloud cover to be surveyed. However, it is now possible to easily expose the state and shape of the earth’s surface as digital 3D visible information with various accuracy technically, economically, and operationally.

Interpretation of landslide topography from contour maps: Since contour lines are data as shown in Figs. 7, 8, and 9, the flexion, density, and spatial distribution pattern of these contour lines are considered to reflect the state of the terrain itself. Therefore, we will understand what shape the actual landslide terrain is, and assume how the landform is drawn as contour lines. Based on this assumption, by carefully observing the contour map, we judge that “such flexion, density, and their spatial distribution pattern are topography that can only be formed by landslides.” This is called deciphering landslide terrain.

Fig. 7

Generate contour lines from point cloud data, judge topographic areas such as water systems, ridges, landslide landforms, and rapid lines by focusing on the bending patterns, and create a topographic classification map

Fig. 8

Various characteristics of landslide topography are summarized and illustrated

1a: Cross-section of a typical Slump type landslide. Upper step: When the slippery surface depth is deep, the main sliding cliff slopes steeply. Lower level: If the slippery surface depth is shallow, the main slide cliff becomes gently inclined. 1b: Decipher the topographic boundary from the contour line, draw the boundary, and make it a landslide terrain. 1c: The distribution pattern of contours corresponding to each of 1a. 2 a,b: Images for real-world vision obtained from a shadow diagram by AW3Dde-ta. 2c: A and b are read by real vision and draw the boundary of the terrain that appears to be a landslide terrain

Fig. 9

Contour lines are generated from the AW3D data 2.5 m DSM (note that it is not 2.5 m DEM), and (1) the water system and ridge line, which are skeletal topographic indicators, are deciphered and drawn. (2) Observe the spatial distribution of contour lines, estimate the landslide topographic area, and map its outline. The landslide terrain consists of a main sliding cliff and a moving body, each decipherable from the spatial pattern of contour bending

Based on this assumption, by carefully observing the contour map, we judge that “such flexion, density, and their spatial distribution pattern are topography that can only be formed by landslides.” This is called deciphering landslide topography.

4.2 Example of Landslide Topography Distribution Map

4.2.1 AW3D-Based Landslide Topography Mapping in Vietnam and Sri Lanka

From 2011 to now, it has implemented three projects in the mountainous regions of Vietnam and Sri Lanka. Those are SATREPS (Development of Landslide Risk Assessment Technology along Transport Arteries in Viet Nam, 2011–2017, Representative, Kyoji SASSA) and SATREPS (The Project on Development of Early Warning Technology of Rain-induced Rapid and Long-travelling Landslides in Sri Lanka, 2020–2025, Representative, Kazuo KONAGAI).During this period, JICA Grassroots Technical Cooperation Project in Northern Vietnam’s Mountain Province (Capacity Building of Local Community for Slope Disaster Risk Reduction, 2020–2023, Project manager, Toyohiko MIYAGI). In these projects, AW3D data constructed by JAXA/RESTEC/NTT data are used to generate contour lines, decipher them, extract landslide topography, and map them.

As described in the ICL Landslide Teaching Tools (2014) above, this series of tasks uses topographic decoding and mapping techniques to identify the locations inherent in the risk of landslides. In Vietnam, aerial photography could not be purchased freely, and a huge budget was required to conduct new aerial photography of the survey area. However, around that time, the AW3D data (5mDEM) of the title was released and sold, and the price was less than 10% of the cost of aerial photography. This was purchased, and the topographic representation of its contours was evaluated (Dung et al. 2016). After that, the necessary parts are purchased sequentially and diagrammed. Before purchasing, you need to verify the quality of your data. Data quality varies. For example, in places strongly affected by clouds, it is difficult to acquire images repeatedly, which is directly linked to the accuracy of topographic expression.

4.2.2 Mapping Results

Vietnam and Sri Lanka are far apart and have different national conditions. However, the geological characteristics of the mapped areas were generally similar. Both were granite gneiss areas from the Precambrian to the Ordovician period. There was another decisive fact. The distribution of landslide terrain is also “wider and more numerous than expected!” In Japan’s granitic areas, landslide fractures are not very frequent. In the early days of mapping, we expected landslides to occur on granitic land to be exceptional. However, the more I compared the site with the map, the more I deepened my understanding, thinking, “I see; that’s why I say that!” Even in granite areas, landslide topography existed widely and densely (Fig. 10).

Fig. 10

A slope disaster in Lao Cai Province in northwestern Vietnam in Sept. 2004. All large-scale landslides, shallow landslides, and debris flows (Put some wards to Yem 2006)

4.2.3 Wide-Area Landslide Topographic Distribution Map 1/25000

Figure 11 shows an example of a distribution map of a part of the project’s target area. Contour maps of 5 m and 25 m were created for about 60% of the data acquisition area, and this topographic map was deciphered.

Fig. 11

Landslide topographic distribution map in some areas of Lao Cai Province in northwestern Vietnam

As shown in the Index, the data acquisition range was divided into a grid of north-south and east-west 10 km and deciphered sequentially. Interpretation was carried out in the form of repeated prototypes according to the progress of the project, so there were differences in the legend and display for each figure width. In that sense, it is a draft version.

In the upper right (northeast) and part of the northwest of Fig. 12, there are places where a huge landslide topography with an area of several square kilometers is located. On the other hand, in the lower half of the area, small and medium-sized landslide topography develops at a high density. Broadly speaking, the granitic great relief mountain slopes correspond to the former, and the middle relief mountain slopes composed of Paleozoic metamorphic rocks and sedimentary rocks correspond to the latter. Two new findings can be obtained here. (1) In general, in Japan, there is not much distribution of landslide topography on mountainous slopes composed of granites, but here it is extremely dense and widely distributed. (2) There was no recognition that the landslide topography was distributed this way. Depending on the type of analysis data and analysis method, the existence of such a natural phenomenon will be revealed as a completely new fact. By grasping these facts, we can think about what kind of understanding and management of slopes will develop in the future.

Fig. 12

Landslide topography distribution in some areas of Lao Cai Province, Vietnam

As a division of the SATREPS project in Sri Lanka (Landslide Assessment Group), it conducts multiple mapping and surveys in four districts: Aranayake, Kandy, Athweltota and Kotamare. A part of it is introduced in Figs. 13 and 14.

Fig. 13

Landslide topographic distribution map near Kandy in the central highlands of Sri Lanka

Figure 13 shows the generation of contour lines with contour spacing of 5 m from AW3D 2.5 m DSM data for the Kandy area of the central highlands of Sri Lanka, and deciphering the water system, ridgeline, and landslide topography (main slip cliff and moving body) from this contour line. The western part of Fig. 13, a-b is distinguished by a steep ridge extending in the south-north direction. Only a few small landslides can be seen on the northeast side of this steep ridge. On the other hand, at the foot of this ridge, there is a gentle slope that seems like a cliff cone. Large-scale landslide topography with a width of about 0.4 ~ 1 km is concentrated at the upper end of the water system in c, d, e, f, etc. The lower end of f is further subdivided, and debris-flowing topography g can be confirmed. Low mountains and hills spread around the river at the northern end of the figure, but landslide topography collects some areas such as h and i.

5 Toward an Era in which Everyone Can Grasp the Terrain in Three Dimensions

Substantive reading of aerial photographs has long been used as an expert competence. However, it was difficult to share the contents of the reading with others. This is because substantive vision is an image in the brain. This led to the availability of digital 3D information, such as AW3D data, and three-dimensional information could be shared. UAVs can even acquire, process, and visualize data on their own. It is possible to visualize things on the earth’s surface in three dimensions and share their reality with multiple people. In modern times, it may be said that the meaning of aerial photography reading has already been lost. In the first place, to grasp the characteristics of landslide topography and slopes, it is essential to have the ability to create an opinion such as “Observing an unspecified number of slopes, is this a slope with such characteristics?” Even if it is possible to convert wide-area 3D digital information into 3D data at once, it is also true that the empirical knowledge of both sides is necessary to grasp the distribution of instability with different characteristics from that vast area.

If this is the case, the significance of the visual interpretation of aerial photographs will continue to be considered material for inferring the actual state of topographic change.

In the next chapter, starting with the visual interpretation of aerial photographs, we will briefly explain the technology for mapping slope changes, such as landslide topography using precise contour lines, the technology for visualizing topography, especially the microtopography of landslide terrain using DEM, and the technology for sharing microtopographic information that is familiar and problematic using UAVs.

6 Aerial Photo Interpretation and Slope Disaster Risk Mapping Training for Beginners

6.1 Focus and Aim

The method of 3D interpretation of aerial photographs has been practiced for about 50 years. This method is to see two overlapping photos as entities, so there is no need to create the data itself. Taking aerial photographs is often a national undertaking. As end users, we purchase, download, and use it. On the other hand, it requires a unique technique of “using two photographs to see the material”. Each sees a different image with both eyes, and the two images materialize and are observed in the brain. A third party can’t see the observer’s brain. Herein lies the difficulty of reading aerial photographs. Third parties will have no choice but to accept or reject the distribution map of the landslide topographic map prepared by skilled observers as if it were a priori fact. This is the background to deciphering landslide topography, where a masterly performance is born. This suggests there is a limit to the establishment and deployment of technology. Here, we will devise ways to gain a common understanding of the entrance to topographic interpretation by aerial photographs. To understand the reality of topographic interpretation assembled by the brain with a third party, I would like to explain the process from interpretation to diagramming in five steps. Some people may find it troublesome, but you can skip reading.

6.2 Aerial Photo Interpretation “Steps for the Technicalization”

6.2.1 Stage 1: Seeing Aerial Photography As a Reality

Observing the terrain three-dimensionally: The terrain is three-dimensional. Therefore, it is better to look at it three-dimensionally to get closer to the actual situation. Then, how much range do you think humans see as three-dimensional? In technology, to see an object three-dimensionally, it is necessary to use (1) “parallax” to look at the object from different angles with both eyes. (2) The difference in parallax obtained by these two eyes is synthesized in the brain and transformed into a three-dimension. Humans can perceive objects three-dimensionally within 10 m at most. In aerial photo decipherment, you look at two pictures with two eyes. A camera with a central projection took each photo from the sky. Adjacent images are taken so that they overlap about 60%, so even if they look similar, there will be a misalignment. It artificially creates parallax. There is a tool called an aerial stereoscope (Fig. 14). It puts two photos side by side, making it easier to see each separately with both eyes. With this tool, you can easily experience stereoscopic images of aerial photographs.

Fig. 14

Three-dimensional observation of aerial photographs of stereo pairs. 1. The close-contact print version of the aerial photograph describes the flight course, scale, shooting time, etc. 2, A photo of a simple stereo mirror and a stereo pair that can emphasize the undulations by about 2 times. 3, Reflective stereo mirror and photo. Pay attention to the distance between the two photos. 4, To facilitate real vision with this stereoscope, the eye is placed on the eyepiece prism from vertically above. The spacing between the two photographs should be fine-tuned to make it easier for the observer to see

Let’s take a look first: Prepare a stereo mirror and two or more sequential aerial photographs. Aerial photographs are marked with the time taken and the code number. First, let’s arrange the photos (Fig. 14-1). This photo was taken closely on a roll film with a scale of 1.15000 and a width of about 24 cm. When magnified about 2–3 times, it has a resolution of about 1 m. Adjacent sequential-numbered photographs can be viewed three-dimensionally with the naked eye (also called the naked eye), a simple stereoscope (Fig. 14-2), or a reflective stereoscope (Fig. 14-3). The overlapping photos are shifted, each viewed with the right and left eyes. Still, the distance between the pupils of both eyes is about 6.5 cm, so shifting the same subject by about 4–6 cm makes it easier to see three dimensions. Since the observation range is very narrow, it will be easier to see the photograph three-dimensionally if you observe it at a distance like a reflective stereo mirror (Fig. 14-4).

If you are a complete beginner, let’s observe it. What do you think? Those who see aerial photography for the first time are almost always surprised. It’s amazing, it looks three-dimensional, it looks like it looks more detailed, it’s like a bird!

6.2.2 Stage 2: Talk About Impressions of “Things” That You See As a Reality

Verbalization of visible things and things: If you look at an aerial photo three-dimensionally using a simple tool, everyone has some image. So, the question is. What did the terrain look like? Tell me in your own words what you see and what you see. If I’m not there, talk to yourself. Then, I can share the image, too. Since I observed the same photos beforehand, I imagine that what you are looking at is “maybe something like this.” This will never be known to a third party.

Of course, what you see will vary from person to person. However, most people recognize unevenness, ridges, valleys, mountain tops, and plains.

You may even notice cliffs and steep slopes. Here, the observer observes the actual terrain three-dimensionally. I notice that.

6.2.3 Stage 3: Focus on the Terrain (this Is the Core)

For example, in the plains, Fig. 15 shows two monochrome aerial photographs of Japan taken around 1980. Within this range, I think it can be easily observed three-dimensionally. In general, the pupil distance between both eyes is about 6.5 cm. Here, the distance between the same points in the left and right images was adjusted to about 4.5 cm to make it easier to observe. You will see three when you look at two photos and realize they are three-dimensional images. The left and right are not three-dimensional, while the center image is three-dimensional and clearer. Your brain observes this situation, so the third party does not know. Observing the terrain in this way is called “substantive visual interpretation”.

Fig. 15

Interpreting the plain

The landforms seen and deciphered as real are arranged on the right as a topographic classification map. At first glance, you can see that there is a plain (smooth land) in this range and a slope (ridge, valley, etc.) on the right (eastern). Roughly speaking, in which direction does the river flow? If you observe the plains carefully, you can also observe river channels, lowlands on both sides of the river that seem like floodplains, and small steps (cliffs). The plain also seems to have some steps. Here, the lowlands along the river are divided into Pt 1, 2, 3, and 4, focusing on the step difference in the lowlands along the river. You may also notice that near the mountains, there are fan-shaped slopes (Paf) that slope gently towards the plains and small plateaus (Pt1) that are even higher.

If you make a diagram of what you read above, you will have made a topographic classification map. The slopes are jumbled compared to the plains, and the terrain looks detailed and complex. On the other hand, even in plains, it is not simply flat and quite complicated. The photographs show more than just the terrain so that you can imagine the correspondence between the topography of the plains and land use. There are no settlements in PT1, which is small and attached to the slope, and Pt4, which seems to be greatly affected by rivers. It seems to indicate the possibility that land use is being carried out in response to the risk and convenience of flood damage. PAF is a terrain called an alluvial fan, a land formed by repeated flooding and debris flows flowing out of the mountains during heavy rains. Settlements are also scattered here. Observing these situations lets you consider the relationship between disasters, people, and topography.

Seeing the slope substantially: What terrain will the slope area be made of? The slopes are complex and messy, but they are places that suffer erosion. The terrain is mainly eroded by rainwater, and its sediment is carried by rivers and deposited in plains and seas. If this is the case, it can be seen that the seemingly complex slope area is in a situation where rainwater and rivers erode the terrain, making it complicated. The water from rainfall becomes surface water and groundwater, which gather and exert an erosion effect. Due to erosion, the valley muscle develops, a water system is formed, and the not-eroded part becomes an elongated ridge. If you take a bird’s-eye view of the slope area, you will notice that the valleys and ridges form a water system. Figure 16 shows a part of the slope area extending to the east of Fig. 15 and shows an image for real vision and its interpretation results. Let’s take a 3D look. Complex slope areas are building a system that drains water like tree branches. The slope area, consisting of small ridges and valleys, is like a treetop, so to speak. In the water drainage system, countless treetops gather, forming large valleys and growing into rivers. The set of small valleys and surrounding ridges is called the valley head. On the ridges, rainwater moves divergently, so the erosion power is not exerted. On the other hand, in the valley, water gathers and exerts great erosion and transportation power. This large difference in action in the system creates a complex and undulating terrain.

Fig. 16

Topographic composition of the place where the slope spreads 1; Crest/ridge system, 2; Stream network 3; Valley side slope, 4; Landslide scarp and the body, 5; River terrace

6.2.4 Stage 4: Observing the Longitudinal Profile from the Ridge to the Valley Floor and Considering the Formation Process

What is the slope condition between the ridge line (ridge) and the mountain stream (valley floor)? The slope between the ridge and valley lines is wide and is named the side slope. I enlarged a part of Fig. 16 in Fig. 17. The ridge is difficult to erode, and the valley floor is where erosion and transportation are extremely intense, including running water. What kind of action works on the valley wall slopes separated by these two landforms? The most important feature of the valley wall slope is the steep slope.

Fig. 17

Close up the valley side slope 1; Crest/ridge system, 2; Stream network, 3; Surface landslide / Slope failure (Red: non-vegetation, Light green: poor vegetation), 4; Vegetation covered side slope, 5; Scree/Talus slope

Figure 17 schematically shows the characteristics of the longitudinal section from the ridge to the valley. Figure 17 divides the valley wall slope into three parts. The red color looks like a collapsed mark on a slope with no vegetation. This is the site where a shallow landslide (surface failure) has recently occurred. The areas shown in yellow-green are noisy with vegetation compared to the surrounding area. Perhaps herbs and shrubs are sticking to it. Other valley wall slopes seem to be covered with forests. In addition, in the lower quarter of the figure, there is a place where you can observe several elongated streaks that lack vegetation. Looking at these planar distributions, it can be said that the valley wall slope is a place where sediment moves toward the valley floor. Sediment movements range from recent occurrences (bare ground) to past occurrences and vegetation recovery, and in some places, debris flow seems to have occurred. In summary, valley wall slopes can be considered as slopes formed by shallow landslides (surface failures) and debris flows that occur repeatedly over a long period of time at various times, scales, and intensities.

Figure 18 shows an aerial view of a debris flow in Hiroshima City in 2014. The disaster brought total rainfall to 2000 mm and resulted in 74 casualties caused by numerous debris. The debris flow is seen as long, narrow streaks carving the slope. Looking at individual streaks, we can observe places where vegetation is lacking, which is considered to be the origin of debris flows, mountain streams where bedrock has been exposed by erosion downstream, and gentle slopes due to the accumulation of moving sediment. The affected residential area is located on a gentle slope. The place where the debris flow began to accumulate and diffuse seems to be at the upper end of this gentle slope. It can be judged that the damaged residential area was created on a debris flow fan formed by the accumulation of repeated debris flows in the past. It can be judged that a phenomenon similar to the debris flow that caused the 2014 disaster has occurred repeatedly. The streak-like microtopography in part of Fig. 17 seems to be the lower part of a small debris flow. Rivers quickly transport sediment brought in by debris flows. However, some parts of the river have sediment accumulations, and there was a large-scale sediment supply in the past.

Fig. 18

Detail of Debris flow that occurred in 2014 in Hiroshima City, Japan

6.2.5 Stage 5: Deciphering Landslide-Induced Terrain

In Fig. 16, many sites that are thought to have been created by landslides have also been identified. This kind of thing is called landslide topography. Landslide terrain is a landform formed by a part of a slope causing landslide deformation. This terrain has topographical features that are very different from the phenomenon on the slopes we deciphered in Stage 4. Typical examples are illustrated by Vernes (1978) and others. Landslide topography created by landslide action is a terrain composed of topographic elements such as main sliding cliffs, moving objects, and slippery surfaces. Among them, the slip surface is the boundary between the moving body and the immovable body, and in many cases, it does not appear on the ground surface. Landslide topography varies in size from (1) deforming part of the slope to a huge landslide that changes the distribution of ridges and water systems. (2) Geological conditions are often involved in forming the slip surface. (3) Once destruction occurs, an autonomous destruction process is followed as described later. (4) The time from the first destruction to the end of the landslide itself is about 103 ~ 5 years. It is quite different from the destruction of the ground surface, which fluctuates depending on the topographic conditions of the slope.

Why should landslide landforms, which result from landslide movements, attract attention in slope disaster prevention? There are two main reasons. (1) Since the factors that cause fluctuations in landslide topography are maintained as they are, even if they are the result of landslide fluctuations, there is a possibility that the slope surface liquefaction will become destabilized again if pore water pressure near the slip surface due to a large amount of precipitation, elevation, earthquake motion, etc. occurs. (2) Land that has suffered landslide fluctuations has suffered many changes in physical properties, shape, water, weathering, and other chemical composition. These changes seem to have increased their vulnerability compared to before the landslide movement. In other words, it will be more vulnerable than immutable land. It can be considered that this vulnerability will predispose to future fluctuations, and further weakening due to fluctuations will create a predisposition to the next fluctuation. Considering this, it can be evaluated that the landslide terrain itself is a sloped area with a risk of landslide fluctuations. The idea that chains trigger primary landslide fluctuations to the next is described as “Autonomous Destruction Process of Landslide Landforms” (Figs. 3 and 4 in this paper, Miyagi et al. 2004, Miyagi 2021)”. Figure 19 shows a seismically pumped giant landslide, which is thought to have occurred by expanding upward from a large-scale landslide that had occurred in the past. Figure 20 seems to be an active landslide, but as shown in the image on the right at a time interval of about 25 years, vegetation seems to have recovered if the movement is slow.

Fig. 19

Seismically pumped giant landslides (bottom: Aratosawa landslides caused by the 2008 Iwate Miyagi Inland Earthquake). Above: State before the 2008 earthquake (multiple landslide topography that fluctuated greatly can be confirmed) The C-14 age of the buried trees in this mobile body was measured to be older than 46,000 years

Fig. 20

The Toshichi hot spring large-scale landslide continues to fluctuate vigorously

Aerial view for real vision (1/1.5000 taken in 1976) Google Earth image taken in 2021.

Repeated large-scale landslide fluctuations can cause major changes in the distribution of the water systems and ridgelines that make up the slope area. Figure 8 shows an example. Landslide topographic changes seem to proceed differently from surface failure (shallow slippage).

6.2.6 Stage 6: Find and Map Landslide Terrain Using Aerial Photography

So far, we have observed various features of the terrain (undulations, steps, slopes, longitudinal sections, etc.), starting with the experience of seeing aerial photographs as reality. In Stage 4, we focused on the slope and shape from the ridge to the valley floor. On the slopes of the valley walls, sediment movements such as surface failures and debris flows frequently occurred. I could imagine the possibility that the slopes of the valley walls, which are now covered with vegetation, have been repeatedly subjected to similar sediment movements. Stage 5 focused on large-scale landslide terrain. I realized that large-scale landslides are phenomena that boldly deform microtopography, such as ridges, valleys, and valley wall slopes. As shown in Stage 3 (Fig. 15), the land formed by such a landslide was scattered like a mosaic in the slope area, as shown in Stage 4. By the way, Fig. 15 is a landslide topographic distribution map.

The slope area is probably a mixture of the slope area that can be grasped in the longitudinal section of the ridge-valley wall slope-valley floor and the area of large-scale landslide terrain. For example, suppose this condition is illustrated on a 1:25,000 topographic map. In that case, there will be a limit to the ability to express the evidence of surface failure and debris flow that occurred in the area that can be grasped in the longitudinal section. Moreover, even slopes currently covered with vegetation themselves are slopes where soil and vegetation have been regenerated after slope failures or debris flows have occurred. It will be difficult to read this minute situation in detail from the contour lines of the topographic map.

On the other hand, terrain created by large-scale landslides has features that stand out from the surrounding slopes due to microtopography, such as sliding cliffs, moving objects, and slippery surfaces, as exemplified in Stage 5. In this case, in some cases, microtopography may also be expressed on the contour lines of the topographic map. Figure 21 shows such an example. Think about the landslide terrain, why you can say so, what the evidence is, what kind of drawing you can draw on the topographic map, etc., and draw it on the topographic map on the far right. If the distribution of landslide terrain is shown extensively, it will be a landslide topographic distribution map, and it will be data that suggests slope instability.

Fig. 21

Photograph of landslide topography for real vision and base map for creating landslide topographic distribution map

The terrain formed by a large-scale landslide is a set of steep cliffs (steep slopes that border the surrounding immovable area and the landslide’s moving body, such as main sliding cliffs, side cliffs, and secondary cliffs that occur partially in the moving body), and mobile bodies (A part of moving materials of earth that slide separated by cliffs and stop on the slope).

7 Understanding Landslide Topography Using Contour Lines and Training for Evaluating Slope Instability

7.1 Perspectives and Approaches

By reading aerial photographs, we could visually grasp large-scale landslide topography, shallow landslides (surface failure), debris flows, etc., and map them. However, it was necessary to transcribe it on a topographic map to map it. For this purpose, accurately reading the contour lines is also necessary. In modern times, if you create a GIS-based reading result with position coordinates, it can be converted semi-automatically to contour maps and statistical data. However, it is difficult to automate this part of deciphering aerial photographs in the first place because the judgment proceeds in a three-dimensional form in the brain. On the other hand, contour lines should faithfully reproduce the terrain. It can be said that the image of the brain is visualized with information called contour lines. In recent years, numerical information on the ground surface has progressed, and it has become easy to create contour maps, shadow maps, stereo images, etc., using this information. Such numerical digital information can be point-cloud data or multi-view photogrammetry-based data. Here, we train to identify landslide topography by reading contour maps, and to evaluate slope stability from the distribution of contour lines.

7.2 Reading Data

The data used in this process uses LiDAR data to decipher precise topography (contour lines of about 1 m) to grasp the precise topography. In addition, the data is somewhat coarse (data accuracy of about 2.5-5 m grid). AW3D data captured and maintained by JAXA, RESTEC, and NTT data is used to judge terrain. The digital data used here is DSM (numerical surface model: data on the top surface that covers the earth’s surface such as land, buildings, vegetation, etc.) and DTM (digital terrain model, but it is used in the same way as the DEM digital elevation model). DSM reproduces the ground surface’s altitude by decreasing vegetation and man-made objects). The method of acquiring data can be broadly divided into point cloud data when using laser pulses. Also, when using images (such as SfM), it will be photogrammetric data.

7.3 Stages Using Airborne LiDAR Data

A huge amount of 3D point cloud data obtained using an aircraft-mounted laser measuring device is processed by filtering and other processing and acquired as high-precision 3D digital point cloud data, which is further processed to create contour maps. Extremely high-precision digital data can be obtained, but depending on demand, performing data acquisition work and post-processing, such as filtering, is necessary. Therefore, it is made to order and extremely expensive. Here, using a helicopter-mounted high-precision laser data measurement system (Naka Nihon Air survey SAKURA-1), we verified the extent to which laser pulses pass through the forest cover layer to reproduce microtopography in mangrove forests where cross-sectional surveys of topographic vegetation have been conducted (Fig. 22). The SAKURA-1 is a waveform recording type, laser pulse firing frequency of 100,000 rounds per second, measurement density more than 4/m2. The upper row of Fig. 22 shows the laser point cloud two-dimensional distributed contour line and altitude stage size in a range of 1 m in width, including the survey line. The lower part is a topographic vegetation measurement cross-section. They correspond well with each other, and high-precision laser surveying proved that the microtopography of the forest floor can be accurately grasped even in mangrove forests with a canopy cover density of about 85% (Makabe et al. 2015, Unome 2020).

Fig. 22

Comparison of aerial laser measurement data (acquired in 2006) and microtopography, vegetation measurement cross-section of mangrove forests including Okinawa Mudlobster mounds (Makabe et al. 2015; Unome 2020)

In the case of general measurement accuracy (1/m²), when the tree canopy cover exceeds 80% and forest floor vegetation exists, it is difficult to acquire topographic information with an accuracy of several tens of centimeters even if LiDAR measurement is performed. However, it has great applicability by acquiring data with various accuracy depending on the ground surface characteristics to be measured and the measurement purpose.

During the 2008 Iwate-Miyagi Nairiku Earthquake, 3500 landslides and landslides occurred, including the Arato Landslide (Miyagi et al. 2011). Aerial laser measurements were conducted over a wide area to grasp the details of this slope disaster. Figure 23 is a cut out of a part thereof. The created point cloud density was about 1/m² and used here as a contour map with 2 m intervals.

Fig. 23

LiDAR data of terrain destruction during the 2008 Iwate Miyagi Inland Earthquake. A 2-m contour map created from LiDAR data (left) and a distribution map of landslide topography and slope fluctuations extracted from LiDAR data (right). 1; ridge and valley lines and water bodies, 2; Landslide main sliding cliffs and moving bodies, 3; Newly generated slope failures and debris flows

With this level of accuracy, landslide topography can be easily grasped, and bare ground formed by shallow landslides that destroy valley wall slopes/surface failures can be grasped as subtle bends of contour lines.

7.4 Extracting Landslide Topography with AW3D Data

AW3D is the name of a data group created from Advanced World 3D Map. This public-private partnership project between JAXA, RESTEC, and NTT Data. Using four million global images taken by the Advanced Land Observing Satellite “ALOS” launched by JAXA in January 2006, and supplementing Maxer’s images as appropriate, 3D digital data with 0.5 ~ 5 m accuracy worldwide was constructed, and it has been put into service sequentially since 2014. This is a countermeasure to the fact that conventional aircraft and manual measurements are extremely expensive and impossible to cover the entire world. AW3D is high-precision digital image information. This is converted into DSM data, and land cover information is processed appropriately to convert it into DEM data. Therefore, if it is widely covered with dense vegetation, the terrain implemented by the DEM may not be able to reproduce roughly. When purchasing data, it’s important to understand how accurate the data in that location is. Other satellite images are the same, but each data has different characteristics. Figure 24 shows the distribution of the accuracy of the data acquired in a region. Clouds usually cover humid tropics, and mountainous areas with extremely dense vegetation. Therefore, it is difficult to obtain good-quality data in such areas. You should also understand the process of generating a DEM from DSM.

Fig. 24

Since it is based on the principle of constructing digital 3D data from the parallax of satellite image data, the higher the number of images obtained repeatedly, the higher the accuracy of the data. Depending on the region, a lot of cloud cover, and valid data may not be available

On the other hand, in areas with poor vegetation (e.g., deserts, high mountains, cold regions), the DSM and DEM are almost identical, and extremely reproducible topographic information can be obtained. Figure 25 shows an example of observing landslide topography on cold and dry mountain slopes. As you can see from the Google Earth image of C and the sketch of D, there is a steep slope in the upper right corner of the slope, a slope with a slight slope directly below it, a steep slope with a slight step in front of it, and a gentle slope below it where you can see the invasion of gullies. There is also a small but distinct cliff in the lower left of the figure. Figure 25(a) shows a topographic map depicting 10 m contour lines from AW3D 2.5 m DEM data. (b) shows contour lines deciphering ridges and valley lines, from which the characteristics of contour lines that are thought to be landslides are deciphered. I tried to decipher the microtopography that constitutes the landslide terrain while comparing it with (c and d). Parts of the main and side cliffs can be identified as steep slopes, and landslide movers can be identified as gentle slopes half-wrapped by steep slopes that make up these cliffs. The end of the moving body cannot be grasped like c.

Fig. 25

Contour map generated from AW3D 2.5mDEM data, topographic readability status, etc. (a): 10 m contour map, (b): Landslide topography deciphered from contour lines, ridge lines, valley lines, (c): Google Earth image of almost the same location, D: C topographic situation sketch

7.5 Samples of the Landslide Topographic Area Mapping

We are working on the SATREPS project “Development of Landslide Risk Assessment Technology along Transport Arteries in Viet Nam, 2011-2017, Principal Investigator: Kyoji Sasa “We were the first in the world to map the distribution of landslide topography along Action Route 7 in Vietnam (Dung et al. 2016). Figure 26 shows the generation of contour lines from 5 m DEM to 10 m. As a result of examining the Grand Trusse, it was found that the outline of the landslide terrain can be well grasped even on a 5 m DEM basis. Figure 26 shows the landslide topography and local conditions. From the two photographs in Fig. 26, the vegetation in the actual area was never dense forest.

Fig. 26

Landslide topographic distribution map along National Highway seven in central Vietnam and site conditions. Landslide topography is widely distributed throughout the region. The landslide topography facing the river is large, but the inside of the moving body is subdivided like a number. From the landscape and landslide edge of the entire region, it can be assumed that the flow bed-like landslide fluctuation is repeated (Dung et al. 2016)

We also use AW3D 2.5 m data in Sri Lanka to create a wide-area landslide topographic distribution map. Figure 27 shows that even with AW3D2.5 m data, there is a subtle difference in the contour flexion pattern between DSM and DEM.

Fig. 27

Landslide topographic distribution map in Sri Lanka and contour characteristics of AW3D 2.5 m data

Fig. 27a shows a 20 m contour line generated from DEM data, and (b) shows a 5 m contour line generated from DSM data. (a and b) are enlargements of parts of A and B. The red areas in these figures are where surface collapse has recently occurred. Fig. a shows the rough topography of the area. B does not remove vegetation information, so it seems difficult to understand because there is a lot of noise. However, the reality is that the rough contour lines of A do not accurately represent the terrain. When exploring microtopography, how to handle vegetation information is considered. Here, I decided to read the DSM data of b carefully. The result is the upper right figure.

Today, digital 3D data from all over the world and free visible images such as Google Earth are freely available. However, each has its characteristics, limitations, and points to remember. It can be said that we have entered an era where it is important to understand and master each individuality.

8 Procedure for Preparing a 3D Map to Detect the Landslide Topography

8.1 3D Topographic Map for Landslide Topography Detection

Topographic information calculated from the elevation value of the DTM is called topographic quantity. It is mainly divided into those that represent the inclination of the slope and those that represent the unevenness of the slope, such as curvature and opening.

Combining these multiple topographic quantities makes it possible to express the topography in 3D using a plan view. This is called a 3D topographic map. 3D topographic maps created from high-precision DTMs can accurately reproduce the topography without the vegetation’s influence.

Ikeda (2020) devised a “3D microtopographic map (MT3DM)” to detect landslide topography using 5mDTM. Here, we will introduce the creation procedure.

8.1.1 The 3D Microtopography Map (MT3DM)

Several topographic maps using DTM have already been devised (Toda 2014). Ikeda (2020) followed the basic concept of CS stereographic drawings and devised parameters with better reproducibility of landslide topography. The main advantages are:

In interpretation of the landslide topography, it is important to extract microtopography such as cracks peculiar to landslides. In order to emphasize these microtopographic boundaries, a curvature map is also created from the high-definition DTM, colored in a warm red color, and superimposed on the topographic map with transparency processing.

In order to grasp the specific height difference between the main scarps and the undulations of the landslide-moving body, it is important to express the stage color according to the altitude. Here, an elevation map with colors set according to elevation is superimposed to make it easier to recognize the terrain elevation visually.

In addition, the color tone of each terrain quantity and the threshold value when considering the color tone is also set to emphasise the representation of the landslide microfeature.

8.2 Creating a Three-Dimensional Microtopographic Map (MT3DM)

8.2.1 Software for Processing the MT3DM

MT3DM is created using open-source GIS. QGIS (http://qgis.org/ja/site/) is used.

8.2.2 (1) Setting and Transforming the Coordinate System

As a general rule, the coordinate system of QGIS uses a projected coordinate system, such as a planar rectangular coordinate system or UTM. It does not use a geographic coordinate system. In topographic analysis, unit inconsistencies occur between Z: elevation data (unit: m) and X, Y: plane distance (unit: degrees), and correct results cannot be obtained.

The coordinate system of the project is set in “Project>“, “Properties”, > “Coordinate Reference Systems (CRS)” (Figs. 28, 29, 30 and 31). Search with Filter and set the coordinate system.

Fig. 28

Project coordinates

Fig. 29

Coordinate variation of geographic information data (shp, raster)

Fig. 30

How to add a layer

Fig. 31

Data storage method

If you need to convert vector and raster data, right-click on each layer of data, click Export >Save Features As, enter a File name, select Coordinate Reference System (CRS), and click “OK” to output the data after coordinate conversion.

8.2.3 (2) How to Add Layers

To add layers> from “Layer” to “Add Vector Layer” and “Add Raster Layer” to deploy on QGIS.

8.2.4 (3) How to Save Project Data

Click Save As > Project, enter a File name, and click Save S to save the project data.

8.3 Preparation for 3D Microtopographic Map

• 1. Preparing DTM Data

Prepare DTM (GeoTIFF, etc., hereinafter referred to as DTM raster) in raster data format obtained by aerial laser surveying, etc. In the case of ground data only, it is converted to raster data with GIS software, etc.

• 2. Deploy DTM rasters on QGIS

Deploy the prepared DTM raster on QGIS. Adding layers is described above.

• 3. Extraction of the scope of coverage

Preparation for topographic mapping: carving out the area of interest (landslide terrain). To crop, click “Raster”, > “Extraction” > “Clip Raster by Extent” (Fig. 32). Select the DEM raster of the Input layer and specify the Clipping extent. There are three ways to select a range. Here, select “Use Map Canvas Extent” to be cut out in the area displayed on the canvas, and click “Run” (Fig. 33). The cropped raster is added to the layer.

• 4. Interpolation of rasters:

In places where elevation values are missing, such as water bodies, interpolate from the elevation around the raster data. Click Fill no data value > “Analysis” in “Raster”. Select the desired raster from “...” in the Input layer and enter the save location and name from “...” in the output raster. Click the “Run” button to start interpolation (Figs. 34 and 35).

Fig. 32

Crop target area command

Fig. 33

Setting the crop command of the target area

Fig. 34

Interpolation of raster data

Fig. 35

The left figure is an example of a DTM raster with missing elevation values in the river channel, and is blank. When this is interpolated, the river part is interpolated as shown in the figure on the right

8.4 Creation of 3D Microtopographic Map (MT3DM)

A three-dimensional microtopographic map (MT3DM) is a topographic map that can obtain a three-dimensional 3D effect by superimposing multiple layers created from DTM data. 3D microtopographic maps are created by combining four types of topographic maps. The four types of topographic maps are “Elevation maps, Slope maps and two types of Curvature maps”. Ikeda created a three-dimensional microtopographic map from 5 m DTM, but this paper also introduces a method to create a higher-definition 1 m DTM (Fig. 36).

• 1. Elevation map creation

First, create an elevation map from a DTM raster. Right-click on the extracted DEM raster and click Properties. Symbology’s “Render type” is “Single band pseudocolor” and is classified into five categories to become red-green-blue from the highest altitude. Adjust the number of color scales with “+” and “-”, and click Value and color to set the elevation value and color. The elevation value is set appropriately while looking at the red-green-blue balance of the created altitude chart. To create an experiential experience, it is better to set the color setting to a dark color close to the primary color, as shown in Figs. 37 and 38.

• 2. Slope map creation

Next, create a slope map from the DTM raster. Click “Analysis” > “Raster” > “slope”. Select the DTM raster as the Input layer, enter the save location after outputting the Parameter to slope and the name, and click “Run” (Fig. 39).

Fig. 36

Flow of creating a topographic map (top: 5 m DTM, bottom: 1 m DTM)

Fig. 37

Color setting of the high-stage color map

Fig. 38

Altitude-stage color drawing

Fig. 39

Slope map creation command setting

Adjust the color of the slope map by right-clicking on Properties >Symbology>“ is Single band grey and “Color gradient“ is set to “White to Black” (Fig. 40). Next, adjust the scale (Min, Max) of the color gradient. The default maximum value (Max) is 90°, but, the Max is empirically set to 45°.

Fig. 40

Slope map color setting

Finally, set the transparency of the slope map to 80% in Properties> “Transparency” (Figs. 41 and 42).

Fig. 41

Slope Map transparency setting

Fig. 42

Slope Map 80% transparency

For any other formatting issue please refer to the.

• 3. Curvature map creation

Create two types of Curvature maps. One is the “Curvature map of small and medium topography”, and the other is the “Curvature map of micro topography”. We aimed to visualize unevenness suitable for each terrain scale. The procedure for creating these two types of curvature diagrams is different for 5mDTM and 1mDTM. The 5mDTM case and the 1 mDTM case are explained separately below.

< for 5mDTM>

• Curvature map of small and medium-scale topography

Smoothing DTM raster to create a Curvature map (Fig. 43) > Click “SAGA” > “Raster-filter” > “Gaussian filter”. Select the DEM raster in “Grid” and enter the name of the output file in “Standard Deviation” 3, “Search Mode” in [1] Circle, “Search Radius” in 10, and “Filter Grid”. Click Run to create a smoothed DTM raster.

Fig. 43

Setting of smoothing command

Next, create a “Curvature map of small and medium topography” from the smoothed DTM raster. > Processing Toolbox> click SAGA Terrain Analysis-Morphometry> Slope, Aspect, Curvature. Select the DTM raster smoothed to Elevation, select the DTM raster smoothed to Elevation, Slope Unit and Aspect Unit“ to [1], and enter the name of the output file in “General Curvature” (Fig. 44). Remove “Slope”, “Plan Curvature“, etc. because processing other than “General Curvature” ☑ is not performed. Click Run to create a Curvature map of small and medium topography.

Fig. 44

Curvature drawing command

Next, set the color of the created curvature diagram (Fig. 45). Right-click on the layer’s curvature diagram to open Properties, under Symbology > Render type (Single band pseudocolor) and classify “Color ramp” into four so that + values (convex terrain) are red and -values (concave terrain) are blue. Min is set within −0.05 ~ −0.03, Max is set within 0.03 ~ 0.05. To emphasize ridges and valleys, Min is set to −0.03, and Max is set to 0.03. Finally, set “Properties” > “Transparency” to 20% (Figs. 46 and 47).

• Creation of the curvature map of microtopography

Create a Curvature map (microtopography) from a raster DTM. The procedure is similar to the Curvature map of small and medium topography. Processing Toolbox> SAGA> Terrain Analysis-Morphometry> click Slope, aspect, curvature. Select the DTM raster under Elevation, select the Slope Unit and Aspect Unit [1], and input the output file into Profile Curvature. Remove ☑because processing other than Profile Curvature, such as Slope and Plan Curvature, is not performed. Click Run to create a curvature diagram.

Fig. 45

Color setting of curvature diagram

Fig. 46

Permeability setting of curvature diagram

Fig. 47

Transparency of small and medium-sized terrain 20%

Next, set the color of the created curvature diagram. Right-click on the layer’s curvature diagram to open Properties. 「Symbology」 > 「Render type」to Single band pseudocolor，and Color ramp set「YlOrRd」. Set the scale within “-0.05 ~ −0.03” for Min and “0.03 ~ 0.05” for Max. Here, in order to emphasize the microtopography, Min is set to “-0.03”, Max is set to “0.03”, and finally “Properties” > “Transparency” is set to “20%” (Figs. 48, 49 and 50).

Fig. 48

Color setting of the curvature diagram of microtopography

Fig. 49

Transparency setting of curvature diagram

Fig. 50

Curvature diagram of microtopography (permeability 20%)

8.4.1 Completion of Three-Dimensional Microtopographic Map (MT3DM)

Slope map with transparent processing based on elevation and curvature maps (small and medium topography). A curvature map of microtopography is superimposed in layer order to create a “three-dimensional microtopography map (MT3DM)”.

Figure 52 shows the “3D Microtopographic Map (MT3DM)” created by superimposing four types of topographic maps (Fig. 51). By superimposing contour lines on a three-dimensional microtopographic map, it is possible to express changes in the height of the terrain in an easy-to-understand manner (Figs. 52, 53, 54 and 55).

Fig. 51

Superposition of four types of topographic maps used for three-dimensional microtopography maps (MT3DM)

Fig. 52

3D Map left: 3Dmicrotopographic map, right: superimposed contour lines)

Fig. 53

Comparison of aerial photographs taken in different years (Geospatial Information Authority of Japan, USA-M1073-22, TO748Y-C1-15, TO20035X-C1-3)

Fig. 54

Interpretation results of landslide topography based on a three-dimensional topographic map (prepared from the 5mDEM of the National Geospatial Science)

Fig. 55

Interpretation results of landslide topography by aerial photograph (TO 699Y-C4A-3-4)

8.5 Landslide Topography Decipherment Using 3D Microtopographic Map (MT3DM)

8.5.1 Topographic Decipherment Using 3D Microtopographic Maps (MT3DM)

Landslide topographic detection by aerial photo interpretation is a difficult technique that requires skilled reading skills, as it is a task to read the true shape of the terrain behind vegetation while considering the development process of landslides. In addition, when focusing on the resolution of the topography of aerial photographs, which has been conventionally used for deciphering landslide terrain, there are differences in photographic equipment, scale, shooting season, shooting time, etc., depending on the aerial photograph. Each aerial photograph has large variations in resolution, saturation, shading, vegetation conditions, etc. Therefore, which aerial photograph is used also greatly affects the accuracy of topographic decipherment.

Therefore, we focus on a three-dimensional topographic map created from DTM data. 3D topographic maps that eliminate the influence of vegetation and reproduce the topography itself help to simplify topographic decipherment, which requires skilled skills. The following is an example of landslide topographic decipherment using a three-dimensional microtopographic map.

8.5.2 Interpretation Example of Landslide Topography in the Active Phase

• Overview of landslide topography

The shape of the main sliding cliff is clear. However, due to the collapse occurring in the main sliding cliff, the general horseshoe shape is not shown, but the hoof-shaped shape with both sides entering the mountainside is shown. In the middle ~ lower part of the moving body, many microtopographies such as cracks (steps) and flow marks are developed, and the microtopographic boundary is clear. In addition, where gully intrusion is observed along the flow scar, the collapse topography of the 0-shaped valley and the development of secondary slips are remarkable. Small and medium-sized landslides have also developed markedly at the ends of mobile bodies.

• Comparison of interpretation results of aerial photographs and three-dimensional microtopography maps

The outlines of landslide topography, such as main sliding cliffs and landslide moving objects, are almost the same interpretation. There was a large difference in the reading results regarding the microtopography inside the landslide topography. In particular, many areas are represented in three-dimensional microtopographic maps but cannot be read in aerial photographs due to vegetation. Furthermore, although similar microtopography could be extracted between the two, there were some differences in the shape of the microtopography (cracks, the extension of steep terrain, etc.). Since the microtopography is obscured by vegetation in aerial photographs, it is often necessary to grasp the shape of the microtopography, and there are many variations depending on the person.

8.6 Points to Note and Issues in the 3D Microtopographic Map (MT3DM)

8.6.1 Color Scale Thresholds and Terrain Representation

Regarding the threshold setting when setting the color of the slope and curvature diagrams, the numerical range is important for clearly expressing the landslide topography.

In the case of a slope map, depending on the setting of the maximum slope of the grayscale, there will be a difference in the shade of the color, which affects the topographic representation. When the maximum value of grayscale is increased, the microtopography in 5 m DTM shows a small angle (about 6° for micro terrain with a specific height difference of 1 m, about 11° for 2 m), resulting in pale colors and blurred microtopographic representation. In the 3D microtopographic map, the maximum grayscale value in the slope map is 45°. The maximum value of the color scale range is 0.05 ~ 0.03, and the minimum value is −0.05 ~ −0.03 for the curvature map. It is explained below.

8.6.2 Topographic Volume Analysis Results and Threshold Settings of Landslide Terrain

Figure 56 shows the results of the topographic volume analysis conducted to clarify the characteristic values of landslide landforms. The analysis analysed two cases of landslide landforms (landslide topography in the active phase and landslide landform in the demolition phase) with a large activity difference (Figs. 56 and 58).

Fig. 56

Case 1 (Landslide landforms in the active phase)

The amount of terrain to be analyzed was calculated from 5 m DTM, and the types of topographic quantities were slope and curvature. There were two types of curvature: “curvature of small and medium-sized terrain” and “curvature of micro-terrain”.

The analysis range of the topographic volume characteristic value is the inside of the moving body where the kinetic characteristics of the landslide appear (Figs. 56 and 58), and the analysis results of the topographic volume characteristic values in each case are shown in Figs. 57 and 59. From the analysis results, even if there was a difference in the degree of landslide activity, no clear difference was observed in the frequency distribution of the topographic volume characteristic values, and it was generally within the same numerical range. In other words, the slope was within the range of −0.05 ~ 0.05, and the range of −0.03 ~ 0.03 was concentrated in the range of −0.03 ~ 0.03. Therefore, this numerical range was taken into account when creating a three-dimensional microtopographic map, and the threshold value was used when setting the color.

Fig. 57

Frequency distribution map of each (Case 1)

Fig. 58

Case 2 (Landslide topography in the demolition period)

Fig. 59

Frequency distribution map of each topographic quantity (Case 2)

Fig. 60

3D landslide topography representation with different DTM size data. Left: 1 m DTM, Middle: 5 m DTM, Right:10mDTM

8.6.3 Landslide Topography Representation and DTM Accuracy

The scale of landslide topography varies from several tens of meters to several hundred meters. The accuracy of the topographic representation of the 3D topographic map depends on the DTM accuracy, and it is considered that there is an appropriate DTM accuracy corresponding to the scale of the landslide. Niida pointed out the relationship between DEM accuracy and topographic decipherment, that there is a mesh size that is easy to read depending on the scale, such as the horizontal and specific height of the terrain. In other words, low-resolution DEMs are suitable for deciphering large terrains, and the higher the resolution, the more detailed valleys due to erosion and the linear topography with sudden changes in slope, such as tectonic lines and faults.

The 3D microtopographic maps are created from 5 m DTM. The accuracy of the terrain representation depends on the DTM accuracy, but it is necessary to understand the case in landslide terrain. We show a diagram comparing 3D microtopographic maps prepared from 1 m DTM, 5 m DTM, and 10 m DTM over a range of about 1 km *1 km (Fig. 60).

The figure created from 1 m DTM showed a lot of noise on the ground surface caused by falling rocks and fallen trees. Large-scale landslides can make it difficult to decipher the terrain. On the other hand, in the case of a small-scale landslide, it is considered that microtopography cannot be clearly expressed without the accuracy of 1 m DTM.

It can be judged that the 5 m DTM is suitable for deciphering landslide topography on a scale of several hundred meters or more. Therefore, 5 m DTM is not suitable for deciphering small-scale landslide terrain. The resolution of the 10 m DTM is too low. Although the outline of the landslide topography is legible, the internal microtopography is obscured even in a large landslide of several hundred meters or more. Therefore, although the main topography, such as sliding cliffs, can be read, it is extremely difficult to decipher the microtopography inside the moving body and is unsuitable for detailed landslide topography decipherment.

9 Conclusion

In this report, we summarize the contents of the exchange of opinions among the co-authors on the following points.

Observing objects in three dimensions is first necessary to grasp, classify, and map the tendency of topographic changes on slopes geomorphologically.

In the past, the main observation method was the real interpretation of aerial photographs, but sensing technology has made great progress in the last 10 years. This has led to major changes in topographic observation approaches. The biggest change is that it is now possible to visualize three-dimensional objects easily. There is a huge variety from aerial photography to the iPhone 14 pro.

Among them, the interpretation of aerial photographs was difficult to understand. Since the actual interpretation is a task of further handling images based on experience and knowledge, it is not easy to share information with others. This also lightens the significance of utilization. However, its usefulness has not been lost. It is effective as a means of observation that can easily observe the details of the current terrain.

On the other hand, digital 3D information can present visible data and maps corresponding to various requirements, even if it is only data processing technology. Even without experience in topographic surveys, disaster risk can be mapped semi-mechanically.

Needless to say, to predict the location of slope disasters, it is necessary to appropriately grasp and classify the topographic changes that take place on slopes, and basic knowledge of this is always necessary.

Aerial interpretation will be necessary to understand local disaster risk. At the same time, digital information will be revolutionary as a direct tool for visualizing and mapping disaster risk. It is necessary to master both.

This report summarizes the outline of the actual situation of slope change and its mapping. In the latter half of the report, we proposed a manual to overcome the difficulty of deciphering aerial photographs. At the same time, we summarized in an easy-to-understand manner a method for visualizing topography three-dimensionally using digital 3D data and grasping the fluctuation trend of landslides.